key: cord-0281864-lje6ugag authors: Yurdakul, Celalettin; Zong, Haonan; Bai, Yeran; Cheng, Ji-Xin; Unlu, M. Selim title: Bond-selective interferometric scattering microscopy date: 2021-06-05 journal: nan DOI: 10.1088/1361-6463/ac0b0d sha: 5ca0e773bc1e081d71c5934f0383b1c6b98f7a52 doc_id: 281864 cord_uid: lje6ugag Interferometric scattering microscopy has been a very promising technology for highly sensitive label-free imaging of a broad spectrum of biological nanoparticles from proteins to viruses in a high-throughput manner. Although it can reveal the specimen's size and shape information, the chemical composition is inaccessible in interferometric measurements. Infrared spectroscopic imaging provides chemical specificity based on inherent chemical bond vibrations of specimens but lacks the ability to image and resolve individual nanoparticles due to long infrared wavelengths. Here, we describe a bond-selective interferometric scattering microscope where the mid-infrared induced photothermal signal is detected by a visible beam in a wide-field common-path interferometry configuration. A thin film layered substrate is utilized to reduce the reflected light and provide a reference field for the interferometric detection of the weakly scattered field. A pulsed mid-IR laser is employed to modulate the interferometric signal. Subsequent demodulation via a virtual lock-in camera offers simultaneous chemical information about tens of micro- or nano-particles. The chemical contrast arises from a minute change in the particle's scattered field in consequence of the vibrational absorption at the target molecule. We characterize the system with sub-wavelength polymer beads and highlight biological applications by chemically imaging several microorganisms including Staphylococcus aureus, Escherichia coli, and Candida albicans. A theoretical framework is established to extend bond-selective interferometric scattering microscopy to a broad range of biological micro- and nano-particles. A light microscope -a common term referring to an optical microscope utilizing visible light to detect and visualize very small objects -is an indispensable tool in a biology laboratory to study biological micro-and nano-particles. The size scale of biological particles is a continuum, ranging from 20 nm for the smallest viruses to a few microns for bacteria and cells. The biological particles have enormous utility as well as potential adverse impacts in biotechnology, human health, medicine, and food safety [1, 2] . For example, viruses are the most abundant biological nanoparticles and they have caused deadly outbreaks throughout human history, with the stark and enduring example embodied in the recent global COVID-19 pandemic. Their physical and chemical properties strongly influence their biodistribution and interactions with cells [3] . Optical microscopy of individual biological particles has significantly improved our understanding of biological, chemical, and physical processes in life sciences [4] . While light microscopy is sufficient to observe biological microparticles such as bacteria, its ability to study individual sub-wavelength nanoparticles is limited by their weakly-scattering characteristics due to exceptionally small size and low-refractive-index difference with the surrounding medium. Elastically scattered light intensity induced by the illumination of the nanoparticles scales with the sixth power of the particle size resulting in very low optical contrast which makes them indistinguishable from the background in scattered intensity-based optical imaging and detection techniques. As an indirect method, fluorescence microscopy has been traditionally used to overcome this weak contrast by detecting exogenous or endogenous fluorophores. It has led to several breakthroughs in the super-resolved imaging of nanoscale specimens [5] . Yet, fluorescence labels suffer from phototoxicity, photobleaching, photostability, and saturation. These severely limit observation of the molecular/nanoparticle dynamics at very short time scales. Moreover, fluorescence labeling may perturb sample functionality and structure further emphasizing the need for biological imaging and characterization without any labels and modifications in their natural environment. Therefore, it is highly desirable to advance light microscopy to provide sufficient resolution to study the entire size range of biological processes at the micro-and nano-scale [6] . Furthermore, augmenting light microscopy with compatible characterization techniques to achieve chemical specificity would provide an invaluable capability especially if both microparticles (bacteria) and nanoparticles (viruses) can be studied. Here, we exploit interferometric microscopy to develop such a capability. While the present experimental work focuses on bacteria characterization, we establish a theoretical framework and requirements to study single biological nanoparticles such as viruses. Over the last two decades, wide-field interferometric microscopy techniques have successfully demonstrated high-throughput, sensitive, and fast direct imaging of biological particles down to a single protein level [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17] . These interferometric techniques rely on the coherent detection of the elastically scattered field -rather than its square or intensity -mixing with a strong reference field in a common-path interferometry configuration. Thus, interferometric detection enjoys a reduced size dependence of optical contrast (scaling with the third power of the particle size) and enables a significantly enhanced nanoparticle visibility. Interferometric scattering (iSCAT) [7] microscopy detects nanoparticles captured on a glass coverslip in reflection mode. iSCAT illuminates the sample with a highly coherent light source, that is laser, to increase the interferometric contrast [18] . Similarly, coherent brightfield microscopy (COBRI) uses the laser illumination in transmission mode followed by background attenuation in the pupil plane to reveal high interferometric contrast [19] . Unlike the laser-based interferometric microscopy methods, single-particle interferometric reflectance imaging sensor (SP-IRIS) [20] uses a layered silicon substrate with a partially coherent light source, that is a light-emitting diode (LED). These common-path interferometry based wide-field microscopy techniques have demonstrated several applications for imaging, detection, and counting of proteins [7, 8, 9] , viruses [10, 11, 12, 13] , exosomes [14, 15] , microtubules [16] , and metallic nanoparticles [21, 22, 19] . However, molecular specificity in these interferometric studies is limited to the surface affinity of assays. Vibrational spectroscopic imaging methods have enabled molecular fingerprinting of molecules with high chemical specificity [23] . Infrared absorption and inelastic Raman scattering have been utilized in various optical sensing methods to probe spectroscopic signatures of chemical bonds. Raman scattering microscopy techniques rely on a non-linear scattering process to achieve diffraction-limited-resolution [24, 25] . Giving complementary information with Raman scattering, infrared (IR) absorption is another type of vibrational spectroscopy method that is routinely used in analytical fields. However, IR-based methods have limited applications in resolving nanoparticles since the spatial resolution is typically several micrometers due to long excitation wavelengths in the mid-IR region (2.5-25 µm) [26, 27] . As a tip-based approach, atomic force microscope infrared spectroscopy (AFM-IR) techniques can resolve structures at the nanoscale (20 nm) but its applications are typically limited to dry samples due to contact requirement [28, 29] . Recently developed mid-infrared photothermal (MIP) microscopy has demonstrated optical detection of photothermal effect induced by IR absorption of the specimen using a visible probe beam [30] . In principle, the MIP contrast mechanism is similar to those reported in the photothermal microscopy field [31, 32, 33, 34] . This emerging technique has been recognized widely in the chemical imaging field and continuously evolved in various geometries including scanningbased [30, 35, 36, 37, 38] and wide-field [39, 40, 41, 42, 43, 44] illumination. A wide range of applications spanning from material science to life science has been reported by scanning MIP techniques [45, 46, 47, 48, 49, 50, 51, 52] . Although providing great sensitivity to characterize specimens down to single viruses [51] , the scanning MIP approaches have inherent limitations associated with scanning itself such as inefficient use of IR photons due to IR/visible foci mismatch and low throughput due to the limited imaging speed. Bai et al. [39] has overcome these challenges and demonstrated a wide-field MIP microscopy approach by probing interferometric reflectance change from the sample. To achieve camera-based photothermal detection using very fast complementary metal-oxide-semiconductor (CMOS) cameras, a virtual lock-in camera technique has been introduced. Camera-based MIP systems have been already applied to extend various label-free imaging methods including low-coherence interference microscopy [44] , quantitative phase imaging [40, 41, 42, 43, 53] , and dark-field imaging [54] and led to applications in histopathology and living cells. However, individual nanoparticle detection in wide-field MIP configuration remains still challenging due to the very weak scattered light from nanoparticles. As a consequence, single nanoparticles become invisible under the strong background from the illumination. In this study, we show bond-selective interferometric scattering microscopy for biological nanoparticle fingerprinting. We utilize a layered silicon substrate to increase interferometric contrast by reducing background and enhancing the scattered field [55, 56] . By probing the mid-IR absorption-induced photothermal signal, this study enables chemical specificity beyond the surface affinity. Our technique provides vibrational spectroscopic sample information which is not accessible in the previous wide-field interferometric scattering microscopy approaches. A theoretical framework for the interferometric photothermal contrast mechanism is discussed. The system performance is evaluated using poly (methyl methacrylate) The visible incidence field (E i ) scatters from the sample (E s ) and reflects off the substrate surface (E r ). The mid-IR pump beam absorbed by the nanoparticle causes temperature rise (∆T ) and induces a change of the particle's polarizability (∆α) which is a function of the particle's size and refractive index. The pump is incident on the sample with an oblique angle (θ =∼ 62 • ) close to Brewster's angle to improve the IR transmission and avoid absorption by the OL. (c) Virtual lock-in camera detection. The synchronization and timing are controlled by a pulse generator which externally triggers the pump, probe, and camera. T period of the IR and pump pulses is set to 200 kHz. The camera frame rate is set to 200 Hz. The pump beam is modulated at 100 Hz such that odd frames are IR-on (Hot) and even frames are IR-off (Cold). Time delay t d between the pump and probe pulses is controlled to measure the transient thermal response from particles and maximize the photothermal signal. The schematic of our bond-selective interferometric scattering microscopy is illustrated in figure 1 (a). The green pump beam is frequency-doubled from a femtosecond (fs) laser (Chameleon, Coherent) operating at 80 MHz with 1040 nm wavelength and 100 fs pulse width by the second-harmonic generation (SHG) process using a non-linear crystal. The second-harmonic generated laser pulses have a 520 nm central wavelength with a 9 nm bandwidth, reducing speckles in the images owing to the low temporal coherence length. Before the SHG process, the fs beam is chopped by an acousto-optical modulator (AOM, is translated by a manual XYZ stage with differential adjusters (PT3A/M, Thorlabs). To achieve normal incidence on the sample, we ensure that the green laser is focused at the center of the back-pupil by adjusting two tip/tilt mirrors before the field lens. The normal incidence realizes a radially symmetric point spread function (PSF) [57] . We utilize this fact to fine-tune the mirrors by imaging spherical beads. The epi-illuminated sample is then imaged onto a 5.0 MP monochromatic area-scan CMOS camera (GS3-U3-51S5M-C, FLIR) through a tube lens (L2). A pulsed mid-IR tunable quantum-cascaded laser (QCL, MIRcat, Daylight solutions) illuminates the substrate from the backside at an oblique angle of ∼61 • . The collimated IR beam is focused on the substrate's front surface using a parabolic gold mirror (P1). There are two main reasons for using the oblique IR illumination: (1) it avoids the IR absorption by the objective lens which creates random signal fluctuations and (2) it can improve the IR transmission that enhances the photothermal signal. To increase the IR transmission in the silicon substrate, we utilize Brewster's angle in which p-polarized light reflection reduces to nearly zero. To do so, we topologically rotate the polarization state of the factory default IR beam from s-polarized to p-polarized which has been detailed in [54] . Briefly, we use two 45 • gold mirrors in a periscope configuration in which two mirrors point to each other at a 90 • axial rotation such that the incident beam along the x-axis propagates at the y-axis. The first mirror reflects the light upwards and upon this reflection the polarization state changes from s to p. The second mirror preserves its polarization state relative to the optical axis of the substrate. The backside oblique IR illumination approach overall improves the IR transmission by about four-fold compared to that of the initial s-polarized beam, providing higher IR power on the sample. Our bond-selective wide-field interferometric microscopy utilizes common-path interferometry configuration for highly sensitive and stable coherent detection of scattered light from nanoparticles. It uses a 70 nm thin layered Si 3 N 4 /Si substrate as depicted in figure 1 The nitride film thickness d is nearly ∼ λ Si 3 N 4 /4 where λ Si 3 N 4 is the wavelength of the light propagating in the nitride layer. This specific layer thickness not only reduces the reflected light through destructive interference but also enhances the total scattered light through the constructive self-interference of the forward and backward scattered light [55] . In this configuration, the incident field E i scatters from the sample E s and reflects off from the layered substrate E r with a complex reflection coefficient dictated by thin-film effect. The resulting total driving field E d becomes the coherent sum of the complex incident and reflected fields, The scattered field amplitude |E s | scales with the total driving field E d and particle polarizability α in the dipole limit where the particle size is much smaller than the illumination wavelength [58] . In the dipole approximation (r λ), the polarizability of a spherical nanoparticle is given as follows [59] : where r is the particle's radius, p is the particle's dielectric constant, and m is the medium's dielectric constant. Equation 1 indicates that scattered light intensity is a function of the particle's volume square (V 2 ∝ r 6 ) and the refractive index difference between the particle and its surrounding medium. The volume dependence of the scattered signal has strong implications on the pure scattering intensity-based detection (e.g., dark-field [60] ) of very small nanoparticles since the signal rapidly falls off as a result of r 6 dependence. Nevertheless, the interferometric detection reduces this square dependence to a linear detection of the scattered field such that the measured signal scales with r 3 instead. In the interferometric microscopy, the detector captures the intensity signal from the coherent sum of reference and scattered fields I det = |E r + E s | 2 as follow: The first term denotes the reflected field intensity I r = |E r | 2 , the second term denotes the scattered field intensity I s = |E s | 2 , and the last term denotes the interference signal 2Re{E r E s * }. The phase term cos(θ) denotes the phase difference between complex reflected and scattered fields. This phase term has strong implications on signal constituents as it allows for accurate particle size information [56] , dielectric characteristics [61, 62] , and its axial position [22] . For the sake of brevity, we take the phase term as 1 in our formulation, providing the maximized interference signal [19] . Due to the r 6 dependence of the scattering intensity, the nanoparticles of interest scatter very weakly. Under this approximation, the scattered intensity term becomes negligible compared to the reference intensity. Therefore, the detected signal can be approximated as I det ≈ |E r | 2 +2|E r ||E s |. This enables the coherent detection of the weak scattering signal with a strong reference. The reflected field intensity constitutes the direct component intensity contribution which can be minimized by simple background subtraction. Here, we define interferometric signal (S) as the background (B) subtracted intensity signal as follows: The interferometric detection realizes the linear detection of the scattered field which is proportional to r 3 whereas the scattering intensity is proportional to r 6 . This leads to high signal-to-noise-ratio (SNR) imaging of small scatterers that generate subtle light scattering contrast above the strong background signal. We further define the interferometric image contrast as the ratio of the interferometric signal to the background intensity [63] , The photothermal signal in our system is measured by probing the change of this interferometric signal. The IR absorption increases the temperature at the nanoparticle's vicinity. This temperature rise ∆T induces a change in the particle's size and refractive index depending on the particle's linear thermal-expansion β r = (1/r)(dr/dT) and thermo-optic β n = (1/n)(dn/dT) coefficients. As a result, the nanoparticle's polarizability and scattered field amplitude change whereas the reference field is untouched. This process is referred to the photothermal effect that causes the temperature-dependent signal change. Our formulation, in principle, is similar to the previously reported photothermal signal theory based on the scattered field from medium fluctuations [33, 34] . Using equation 3, we can express the signal difference ∆S between IR-on (S hot ) and IR-off (S cold ) states as follows: where E hot s = E s (T 0 + ∆T ) and E cold s = E s (T 0 ) are respectively the scattered fields at the IR-on and IR-off states with the pre-IR pulse temperature of T 0 . Equation 5 implies that interferometric photothermal signal is detected through the linear detection of the scattering amplitude change with a strong reference field [64] . This has a similar detection principle to that of the interferometric signal (see equation 2). Analogous to equation 4, the photothermal image contrast can be expressed as the interferometric contrast difference at two states. We measure the photothermal signal as the intensity difference at the camera. To generalize the photothermal signal quantification, we define the intensity modulation fraction which can be expressed as the ratio of the photothermal signal to the interferometric signal at the pre-pulse state M P T = ∆S/S cold . By plugging equation 1, the modulation fraction can be calculated as follows [51] : The equations above indicate that the photothermal signal contrast scales with the modulation fraction and interferometric contrast. Therefore, a reference signal reduction through either substrate engineering [61] or pupil engineering [55, 65, 19] could enhance the photothermal signal visibility. The photothermal modulation is typically in negative five/six orders of magnitude for a given 1K temperature rise. For example, PMMA has a linear thermal expansion coefficient of β r = 90 × 10 -6 K -1 [66] and thermo-optic coefficient of n PMMA β n = -1.1 × 10 -4 K -1 [67] with a refractive index of n PMMA = 1.49 [68] at 520 nm wavelength. The modulation fraction for a PMMA bead surrounded in the air then becomes nearly 0.01% at ∆T = 2K. To measure the photothermal signal in the wide-field system, we employ the virtual lock-in camera detection reported in [39] . As shown in figure 1(c) , while a pulse train of visible beam continuously illuminates the sample, the IR beam simultaneously illuminates the sample at every other camera frame. In other words, the IR beam is turned on and off at the subsequent frames. The IR pulse period of 5 µs provides enough time to reach steady-state cooling for nanoparticles which have typically fast cooling time smaller than 1 µs [54] . We refer the IR-on and IR-off frames to "hot" and "cold", respectively. The photothermal image is then obtained by taking the difference image, subtracting the cold frames from the hot frames. then temporally stored in the memory and directly averaged into a single 2D array at their corresponding data arrays as "hot" and "cold", respectively. A large number of frame averaging is required to obtain high SNR that can distinguish the photothermal signal from the background noise. We typically acquire 10000 frames in total, that is, 5000 hot and 5000 cold and average them in real-time. This computationally efficient image acquisition approach significantly reduces the memory and space requirement for a large number of frames (1)). In other words, only two averaged hot and cold images (a few MBs) are saved on the disk at the end of each photothermal image acquisition, instead of saving gigs of image data to average at the post-process. We note that all images can also be saved for SNR characterization experiments. was diluted with deionized water by 1:10, followed by a spin coating on the nitride substrate. In bacteria experiments, we used two strains of bacteria S. aureus ATCC 6538 and E. coli This provided nearly 30 µm × 60 µm of FOV for the photothermal detection. The IR focus can be adjusted to a more circular spot by defocusing the parabolic mirror's position, but the defocused IR focus could reduce the IR intensity and photothermal signal. For the sake of proof-of-concept experiments, we kept the IR focus smaller to achieve a high-SNR signal. Note that, we cropped the rectangular image FOVs into squares to obtain symmetric data visualization throughout the manuscript. Our photothermal microscopy in practice can achieve shot-noise-limited sensitivity. The noise-floor in the interferometric signal is dominated by the photon shot noise [63] . However, the wide-field laser illumination creates heterogeneous background across the FOV. The interferometric contrast from nanoscale samples can be buried under such heterogeneity. Luckily, photothermal detection is immune to those variations, providing nearly background-free images [31] . Owing to the virtual lock-in camera detection described above, the interferometric We experimentally evaluated the noise floor in our imaging system. We measured the noise in differential frames of the same FOV captured at a various number of collected photons (P ). This minimized the background fluctuations due to the nonuniform laser illumination. Figure 2 (b) shows our noise calculations in the logarithmic scale. The fitted curve as a function of P was αP 0.498 with a constant scaling factor of α. The exponent of the curve was very close to the theoretical value of 0.5. This indicated that our camera grants the shot-noise-limited detection for sufficiently saturated frames. The modulation fraction described in equation 7 was in the orders of the camera's shot-noise limited contrast which is defined by 1/ √ P. Averaging of shot-noise-limited image frames can reduce the noise floor by a factor of the square root of the number of frames N. Therefore, multiple difference images were averaged in practice to obtain the photothermal signal with good data fidelity. The PMMA film was a great experimental example to show SNR improvement via frame averaging since a single photothermal frame had an SNR of 1.2. The SNR was calculated by dividing the maximum signal by the background standard deviation outside the IR spot. As shown in figure 2(c), the experimental image SNR curve which scales with αN 0.505 was consistent with the theoretical curve (αN 0.5 ). This indicated that the photothermal SNR can be greatly improved by frame averaging with minimal mechanical noise during the acquisition. Furthermore, the SNR in shot-noise-limited measurements can be further improved by using CMOS cameras with large pixel-well-depth recently used in interferometry measurements [44] . The use of these large pixel-well-depth cameras at high-frame rates would enable ultra-fast single nanoparticle fingerprinting. Next, we perform proof-of-principle experiments on 500 nm PMMA nanospheres to demonstrate photothermal imaging of individual particles. The PMMA beads have a refractive index similar to biological nanoparticles, having similar particle polarizability for a given size (see equation 1). Therefore, the interferometric contrast from PMMA beads is very close to those from the same size viruses or bacteria. Figure 3 (a) shows the interferometric image from a cropped FOV, showing nearly a contrast level of 2. To obtain chemical imaging of these beads, we target the carboxyl group around the C=O stretching absorption peak at 1729 cm -1 . The resonance photothermal image has an SNR of 9. When the IR wavelength is tuned to off-resonance at 1600 cm -1 , we observe no photothermal contrast. As we expected, the background heterogeneity due to laser illumination is greatly reduced in the photothermal images. This shows consistency with the differential images obtained under interferometric scattering microscopy [7] . Furthermore, our photothermal measurements According to Rayleigh's resolution definition, the first minimum of the cross-section profile across the 500 nm beads coincides with the peak value of around 500 nm. Under Gaussian object approximation [69] , the deconvolved PSF becomes 357 nm which is close to λ/2NA. To demonstrate bond-selective interferometric imaging of individual biological nanoparticles, we study bacteria and fungi as testing models. Figure 4 shows multispectral imaging of spherical S. aureus bacteria that are directly immobilized on the substrate (see Meth- The upper right arrow indicates that the negative interferometric contrast from S. aureus due to its size appears to be negative in the MIP images. IR powers: 12.4 mW @ 1650 cm -1 , 9 mW @ 1549 cm -1 , and 6.05 mW @ 1729 cm -1 . Photothermal image acquisition time: 50 s (5000 frames). Scale bars: 10 µm ods section). The interferometric image clearly shows their size variation across the FOV since the interferometric contrast is related to the particle's size. To target major protein absorption peak in amide I band, the IR wavelength is tuned to 1650 cm -1 . The obtained photothermal image has a strong photothermal contrast with an SNR of about 60, indicating the rich protein content in S. aureus cells. Furthermore, amide II band at 1549 cm -1 and off-resonance at 1729 cm -1 show the distinctive spectroscopic imaging at the same FOV. The weak-contrast at off-resonance wavelength comes from the residual IR absorption in the vibrational fingerprint region [50] . Owing to the chemical specificity, photothermal detection is immune to the scattered signal from the unspecific immobilized particles. The interferometric particle signal indicated by the left arrow in figure 4 (a) disappears in the photothermal images, showing no contrast. The dark contrast particles (see e.g., the middle arrow), seem to be associated with unwanted signals based on the one-to-one comparison. However, the dark contrast particle indicated by the right arrow shows photothermal contrast. This result further emphasizes the chemical specificity of our technique which was inaccessible in the conventional interferometric microscopy. We also point out that the photothermal contrast could have a positive or negative sign. This stems from the fact that thermo-optic and thermal expansion coefficients are typically in opposite signs due to the particle density reduction [36] , counteracting each other according to the superposition in equation 7. The overall photothermal effect on the signal strongly depends on the particle's size [35] . Either one of the coefficients could be dominant or they can cancel each other. Thus, some materials of a certain size could be invisible in this technique. 12.4 mW @ 1650 cm -1 , 9 mW @ 1549 cm -1 , and 6.05 mW @ 1729 cm -1 . Photothermal image acquisition time: 50 s (5000 frames). Scale bars: 10 µm Next, we image rod-shape E. coli bacteria in the fingerprint region as shown in figure 5 . Similar to the S. aureus cells, the multispectral images are acquired at amide I (1650 cm -1 ) and amide II (1549 cm -1 ) bands and off-resonance bond (1729 cm -1 ). The amide I band shows the strongest photothermal signal with an SNR of about 22, which is three-fold less than the S. aureus cells. Nevertheless, E. coli photothermal images also exhibit distinctive signal levels at the amide II band and off-resonance bond which has much weaker photothermal contrast. We also observe that one of the E. coli cells have negative photothermal contrast although its interferometric image contrast is positive. Interestingly, another E. coli with negative interferometric contrast has a positive photothermal contrast. These results indicate that photothermal signal signs of the same particle types could be different even for similar interferometric contrast levels due to the differences in their size and shape. Together, these two distinctive bacteria results show the promise of this bond-selective interferometric scattering microscopy for high-throughput and multiplex differentiation of diverse bacteria populations. We further investigate the capability of our technique on micron-scale single cells. We image C. albicans fungi that have oval morphology. Since these cells are several microns in diameter larger than diffraction-limited resolution, the temporal coherence of the laser source hinders visualizing inside the cells due to the speckle noise. To address this limitation, we replace the green laser illumination with a partially coherent blue light-emitting diode (LED). This greatly reduces the speckle noise as shown in figure 6 . The LED illumination is achieved in Köhler configuration which provides source-free and highly uniform sample illumination. We note that the frame rate in this experiments is set to 1250 frame/s using a fast CMOS camera (IL5, Fastec Imaging). To measure the protein distribution inside the cells, we tune the IR wavelength to 1650 cm -1 at amide I band and 1450 cm -1 at the amide III band. The strong signal from protein bonds is obtained whereas the off-resonance image shows no contrast. To avoid the lipid absorption peak, the off-resonance wavelength is set to 1800 cm -1 in contrast to bacteria samples. Furthermore, we target the phospholipid ester In summary, we report the bond-selective interferometric scattering microscopy that enables fingerprinting individual nanoparticles captured on the layered substrate. We utilize 70 nm Si 3 N 4 /Si substrate to increase interferometric contrast by reducing the reference background. The analytical framework of the interferometric photothermal signal mechanism is formulated. A direct relation between the interferometric contrast and the photothermal signal has been discussed. We demonstrate proof-of-principle biological experiments on S. aureus and E. coli bacteria and C. albicans fungi. Our technique provides high-throughput molecular information beyond affinity-specific molecular specificity without any labeling in the vibrational fingerprint region. Moreover, this direct chemical imaging method could open up exciting possibilities for the functional analysis of biological nanoparticles at a single-particle level. Our study is limited to wavelength scale nanoparticles since photothermal imaging of sub-100 nm particles requires a much shorter pump and probe pulses as a consequence of the direct relationship between heat dissipation and particle size. As substantiated by the experimental results and theoretical framework, we demonstrate that bond-selective interferometric microscopy can achieve chemical sensitivity for a broad range of nanoparticles currently studied by interferometric techniques. Although our mid-IR photothermal technique has been demonstrated on the IRIS configuration with nitride-coated silicon substrates, it can be easily expanded to other interferometric microscopy approaches using transparent substrates including iSCAT and COBRI. In these configurations, standard glass substrates must be replaced by mid-IR transparent microscope slides such as CaF 2 . It should be noted that our bond-selective interferometric scattering microscopy method can be readily applied to the recent enhancements in the common-path interferometric microscopy techniques such as high-resolution with computational microscopy [57] , enhanced visibility with pupil engineering [55, 65, 19] , nanoscale 3D localization with defocus signal [22] , and high precision molecular mass quantification [8] . Such improvements advance the mid-IR photothermal microscopy field towards the chemical nanoscopy of individual biological nanoparticles including viruses, exosomes, and proteins. Overall, our refined bond-selective interferometric microscopy has a great promise for high-resolution and high-throughput spectroscopic imaging of a broad size range of biological nanoparticles without any labeling. Engineering exosomes as refined biological nanoplatforms for drug delivery Food-borne diseases-the challenges of 20 years ago still persist while new ones continue to emerge Biodistribution, pharmacokinetics, and blood compatibility of native and pegylated tobacco mosaic virus nano-rods and -spheres in mice Biological nanoparticles and their influence on organisms Super-resolution microscopy demystified Emerging applications of label-free optical biosensors Label-free, all-optical detection, imaging, and tracking of a single protein Quantitative mass imaging of single biological macromolecules Direct optical sensing of single unlabelled proteins and super-resolution imaging of their binding sites High-speed nanoscopic tracking of the position and orientation of a single virus High-throughput detection and sizing of individual low-index nanoparticles and viruses for pathogen identification Digital sensing and sizing of vesicular stomatitis virus pseudotypes in complex media: A model for ebola and marburg detection Coherent brightfield microscopy provides the spatiotemporal resolution to study early stage viral infection in live cells Digital detection of exosomes by interferometric imaging Label-free detection of nanoparticles using depth scanning correlation interferometric microscopy Label-free imaging of microtubules with sub-nm precision using interferometric scattering microscopy Highly sensitive and label-free digital detection of whole cell e. coli with interferometric reflectance imaging Interferometric scattering microscopy and its combination with single-molecule fluorescence imaging High-speed imaging and tracking of very small single nanoparticles by contrast enhanced microscopy Interferometric reflectance imaging sensor (iris)-a platform technology for multiplexed diagnostics and digital detection Digital microarrays: Single-molecule readout with interferometric detection of plasmonic nanorod labels Interferometric scattering microscopy reveals microsecond nanoscopic protein motion on a live cell membrane Vibrational spectroscopic imaging of living systems: An emerging platform for biology and medicine Raman microscopy: developments and applications Label-free biomedical imaging with high sensitivity by stimulated raman scattering microscopy Using fourier transform IR spectroscopy to analyze biological materials Fourier transform infrared vibrational spectroscopic imaging: Integrating microscopy and molecular recognition AFM-IR: Combining atomic force microscopy and infrared spectroscopy for nanoscale chemical characterization Nano-FTIR absorption spectroscopy of molecular fingerprints at 20 nm spatial resolution Depth-resolved mid-infrared photothermal imaging of living cells and organisms with submicrometer spatial resolution Abdelhamid Maali, Brahim Lounis, and Michel Orrit. Photothermal imaging of nanometer-sized metal particles among scatterers Room-temperature detection of a single molecule's absorption by photothermal contrast Photothermal heterodyne imaging of individual metallic nanoparticles: Theory versus experiment Gaussian beam photothermal single particle microscopy Super-resolution far-field infrared imaging by photothermal heterodyne imaging Infrared photothermal heterodyne imaging: Contrast mechanism and detection limits Approaches to mid-infrared, super-resolution imaging and spectroscopy Phase-sensitive lock-in detection for highcontrast mid-infrared photothermal imaging with sub-diffraction limited resolution Ultrafast chemical imaging by widefield photothermal sensing of infrared absorption Bond-selective transient phase imaging via sensing of the infrared photothermal effect Quantitative phase imaging with molecular vibrational sensitivity Labelfree biochemical quantitative phase imaging with mid-infrared photothermal effect All-digital histopathology by infrared-optical hybrid microscopy Bulk to nanometer-scale infrared spectroscopy of pharmaceutical dry powder aerosols Super-resolution infrared imaging of polymorphic amyloid aggregates directly in neurons Subdiffraction infrared imaging of mixed cation perovskites: Probing local cation heterogeneities Suppressing cation migration in triple-cation lead halide perovskites Mid-infrared photothermal imaging of active pharmaceutical ingredients at submicrometer spatial resolution Fingerprinting a living cell by raman integrated mid-infrared photothermal microscopy Vibrational spectroscopic detection of a single virus by mid-infrared photothermal microscopy Labelfree imaging of fibroblast membrane interfaces and protein signatures with vibrational infrared photothermal and phase signals Adaptive dynamic range shift (adrift) quantitative phase imaging Background-suppressed high-throughput mid-infrared photothermal microscopy via pupil engineering Pupil function engineering for enhanced nanoparticle visibility in wide-field interferometric microscopy Quantitative interferometric reflectance imaging for the detection and measurement of biological nanoparticles High-throughput, high-resolution interferometric light microscopy of biological nanoparticles Principles of Nano-Optics Light scattering by small particles Three dimensional orientational imaging of nanoparticles with darkfield microscopy Physical modeling of interference enhanced imaging and characterization of single nanoparticles Nanoparticle classification in wide-field interferometric microscopy by supervised learning from model Interferometric scattering microscopy: Seeing single nanoparticles and molecules via rayleigh scattering Quantitative imaging of single light-absorbing nanoparticles by widefield interferometric photothermal microscopy Label-free single-molecule imaging with numerical-aperture-shaped interferometric scattering microscopy Physical properties of polymers handbook Temperature dependence of refractive characteristics of optical plastics Spectral and angular shaping of infrared radiation in a polymer resonator with molecular vibrational modes Speckle correlation resolution enhancement of wide-field fluorescence imaging We thank Dr. Meng Zhang for providing bacteria samples. This work was supported by the National Institutes of Health R35GM126223, R44EB027018, and R42CA224844 to J.X.C.